A studio and a lab by Halley Young — an artist-researcher whose medium is the feedback loop between human taste and machine possibility. I write music and poems, and then build strange computational instruments to ask better questions about why a work feels alive. The research is a human-experience-centered inquiry: it starts from what a song feels like, not from the technology. AI as instrument, never author.
Start by Listening
Before any of the research or the tools, there is the music. I'm a composer first — everything I build with AI grows out of years of writing songs, setting poems, and chasing the feeling of a piece clicking into place. So press play before you read another word.
AI should give you more selves, not fewer. I have never wanted to automate the artist away — I want to hand you new instruments, strange mirrors, and counter-egos that send you back to your own work more alive than you left it.
The fear that creativity and AI are at war is real, and I take it seriously. But I think it comes from a narrow use of these tools — as replacements. Used differently, AI becomes an instrument: a mirror, a counter-ego, a sketchbook for impossible variations. A model can generate a hundred options; only an artist can recognize a direction. The judgment, the risk, the meaning — those stay with you. Everything on this site is an attempt to prove that, in public.
The Through-Line: A Telescope Pointed at Creativity
Creativity has mostly been studied from one of two sides. Artists describe it from the inside, as something ineffable. Scientists model it from the outside, as something to be reduced. I'm taking a third path — a human-experience-centered inquiry that treats my own catalog as a laboratory. Every song, label, persona, and proof is an experiment, and my two research questions are the measuring instruments: what makes a song work (does it land — is it actually enjoyed?) and how do we avoid repetition (is it genuinely diverse, or has it collapsed into one voice?).
So as I expand the music — a deepening catalog now, and increasingly organized into distinct labels, each its own aesthetic — I'm not just making more art. Think of each distinct body of work as a coordinate in the space of what people find creative and moving, tuned for both human enjoyment and diversity. The more of that space I sample, the better the map. AI lets me reach a breadth no single composer could cover by hand; human enjoyment keeps every point on that map honest.
The payoff is exactly that map. By deliberately producing the widest possible variety of work that still lands with real listeners, I'm tracing the boundary between novelty that feels alive and novelty that feels dead. That boundary is, I believe, where creativity actually lives — and the whole body of work here is a telescope I'm building to see it more clearly. Understanding creativity, it turns out, may require producing it at a scale and diversity only a human–AI collaboration can reach.
The Questions I'm Chasing
I'm a researcher (PhD in Computer Science, University of Pennsylvania; now at Microsoft Research), but the questions that drive me are the ones any artist would ask. Here are the two I keep returning to — and a few others I'm exploring.
What makes a song work?
Why does one progression give you chills and a nearly identical one fall flat? I try to formalize the felt logic of music — and, just as importantly, to map exactly where that logic provably escapes formalization. Some of what makes art move us is, in a precise mathematical sense, undecidable. That's not a defeat; it's a discovery about why art stays alive.
When I asked ChatGPT for 250 poems, 249 came back echoing John Keats — without my ever naming a poet, era, or style. Models quietly collapse toward a single voice, and I treat that sameness as an aesthetic failure, not just a technical one. Rather than asking vaguely for "more variety," my work lets you name the dimensions of difference you actually care about — rhyme, meter, syntax, form, gesture — and steers the model along those.
Translating a poem into music, an image into a story — not to imitate the artist, but to give them a "non-human comrade" (as poet Charles Bernstein described one of my homages to his work). Augmentation, never replacement.
Ableton is Turing-complete — the same, in a deep sense, as our most powerful computers. I treat that fact as an aesthetic, not a trick: sonification, live performance, and the strange unpredictability that rigor makes possible.
What if your inner narrative could blossom into a field of possible directions that never competes with the work of your own two hands? Give me an "identity statement" — a two-word stereotype you'd like to reclaim ("Swing Composer") or a multi-paragraph account of who you are as an artist — and I unfold it into studies across poetry, music, narrative, and image. Not a replacement voice: a set of mirrors and provocations. Some land uncannily close; some are wrong in useful ways; some point at a direction you'll want to chase by hand. The goal isn't automatic art — it's creative self-knowledge.
No artist should have to feel comfortable handing a computer scientist their life's work. An identity statement is already enough to generate something worthwhile. But if you want to go further, give me a manifesto — a statement of your artistic beliefs — and I'll explore "alternative personae" who could share your convictions yet are decidedly not you. A way to meet the other artists living inside your own values.
Art is a conversation — between old and new, but also between media. My cross-medium homages act as counter-egos: they turn a poem into a living visual narrative, set a non-songwriting musician to song, or derive a short story from a painting. Not "more DALL·E art," but visual poems that draw on and transcend both traditions. When Charles Bernstein called my homage to his work a "counter-ego or non-human comrade," it named exactly what I'm after.
I like to show how mathematics and computation make artistic practice more magical, not less. If you have a theory about your own art, I'd love to help you formalize it in a language mathematicians and computer scientists would recognize — and then watch what the formalization reveals.
A song that almost works. A model that keeps repeating itself. A poetic voice you want to push somewhere new. A creative identity you want to try on. A theory of your own art you'd like to formalize. A medium you want to step into without losing yourself. Send it and we'll build an experiment around it.
